TY - JOUR
T1 - Modeling clinical assessor intervariability using deep hypersphere encoder-decoder networks
AU - van der Putten, Joost
AU - van der Sommen, Fons
AU - de Groof, Jeroen
AU - Struyvenberg, Maarten
AU - Zinger, Svitlana
AU - Curvers, Wouter
AU - Schoon, Erik
AU - Bergman, Jacques
AU - de With, Peter H.N.
PY - 2020/7/1
Y1 - 2020/7/1
N2 - In medical imaging, a proper gold-standard ground truth as, e.g., annotated segmentations by assessors or experts is lacking or only scarcely available and suffers from large intervariability in those segmentations. Most state-of-the-art segmentation models do not take inter-observer variability into account and are fully deterministic in nature. In this work, we propose hypersphere encoder–decoder networks in combination with dynamic leaky ReLUs, as a new method to explicitly incorporate inter-observer variability into a segmentation model. With this model, we can then generate multiple proposals based on the inter-observer agreement. As a result, the output segmentations of the proposed model can be tuned to typical margins inherent to the ambiguity in the data. For experimental validation, we provide a proof of concept on a toy data set as well as show improved segmentation results on two medical data sets. The proposed method has several advantages over current state-of-the-art segmentation models such as interpretability in the uncertainty of segmentation borders. Experiments with a medical localization problem show that it offers improved biopsy localizations, which are on average 12% closer to the optimal biopsy location.
AB - In medical imaging, a proper gold-standard ground truth as, e.g., annotated segmentations by assessors or experts is lacking or only scarcely available and suffers from large intervariability in those segmentations. Most state-of-the-art segmentation models do not take inter-observer variability into account and are fully deterministic in nature. In this work, we propose hypersphere encoder–decoder networks in combination with dynamic leaky ReLUs, as a new method to explicitly incorporate inter-observer variability into a segmentation model. With this model, we can then generate multiple proposals based on the inter-observer agreement. As a result, the output segmentations of the proposed model can be tuned to typical margins inherent to the ambiguity in the data. For experimental validation, we provide a proof of concept on a toy data set as well as show improved segmentation results on two medical data sets. The proposed method has several advantages over current state-of-the-art segmentation models such as interpretability in the uncertainty of segmentation borders. Experiments with a medical localization problem show that it offers improved biopsy localizations, which are on average 12% closer to the optimal biopsy location.
KW - Deep learning
KW - Intervariability modeling
KW - Localization
KW - segmentation
UR - http://www.scopus.com/inward/record.url?scp=85075463227&partnerID=8YFLogxK
U2 - 10.1007/s00521-019-04607-w
DO - 10.1007/s00521-019-04607-w
M3 - Article
SN - 0941-0643
VL - 32
SP - 10705
EP - 10717
JO - Neural Computing and Applications
JF - Neural Computing and Applications
IS - 14
ER -